Noor's Newsletter: Issue #8

This isn’t a recap of the weeks—it’s an attempt to understand the forces reshaping how we live, govern, and evolve.

From Potential to Practice: The Hard Work of Implementation Continues

A decade ago, I entered the healthcare and life sciences world from a pure AI background. Over the past 10 years, I've witnessed transformative technologies reshape this space, from the early days of generative AI to today, where we see drugs designed by AI in clinical trials and AI helping make precision medicine a reality. Having worked on startups shaping this space, I've also seen how slow and painful the transition from lab to clinic can be. The news from the past two weeks offers a real-world illustration of this tension; the challenge is no longer a lack of innovation, but the friction of real-world implementation. The entire ecosystem—from regulators and governments to startups and incumbent giants—is now grappling with the complex challenges of deploying AI at scale.

The frontier of innovation is no longer a research paper; it's the difficult work of integration. A feature in Business Insider this week explores the "messiest relationship in healthcare," detailing the complex dance between AI scribe startups like Abridge and EHR giants like Epic (which runs over 40% of hospitals' EHR systems). This isn't just a story about two companies; it's a microcosm of the entire industry's challenge: how do you embed nimble, powerful AI tools into the rigid, legacy infrastructure where only a few companies hold the keys? How do you support innovation and ensure patients get the best technology when the market is governed by the few?

Governments are now stepping in to build frameworks for this new reality, though some are better equipped than others. In the UK, which has a unique advantage in the space, the government has announced a new commission specifically to accelerate the NHS's use of AI. While this isn't the first commission of its kind—and past efforts have had limited success—it is a direct acknowledgement that adoption requires a concerted strategy to overcome institutional inertia. This move is undoubtedly spurred by increasing pressure from industry leaders, with companies like Merck and AstraZeneca pulling out from the UK, as we mentioned in the previous issue, and Eli Lilly's CEO this week calling the UK the "worst major country in Europe" for drug prices and access. The UK holds some of the world's richest health data and brightest minds, but without the right policies for innovation, implementation, and reimbursement, it risks falling behind.

On the other side of the pond, as the technology evolves at lightning speed, regulators are trying to catch up, with defense and security as major concerns. In a significant move, California's governor has signed a new law requiring AI safety disclosures. This isn't about stifling innovation; it's about building the trust necessary for it to flourish in high-stakes environments. By setting a baseline for transparency, regulators are creating the stable, predictable conditions that both innovators and healthcare providers need to move forward.

• Noor

Interesting Things in Research and Beyond

One of the most promising frontiers in biology is the application of language models to the "language" of chemistry and proteins. A new paper on  bioRxiv caught my eye this week, exploring how protein language models can be used to understand the evolution of viruses. By treating viral proteins as sequences of text, researchers are able to train models that can predict how a virus might mutate to evade the immune system, and even design new antibody therapies to neutralize it.

Signals to Watch

• The "Picks and Shovels" Investment: Private equity firm Kedaara Capital has invested $240 million in Axtria, a data analytics and software provider for the life sciences industry, signaling strong investor interest in the infrastructure that makes the sector more efficient.
• The New Economics of AI: China's DeepSeek has revealed that its competitive AI model was now trained for just $294,000, this figure drastically undercuts the massive training budgets often discussed in the West, signaling a new focus on capital efficiency. It suggests a Moore's Law-style dynamic is taking hold for AI, where algorithmic and hardware optimizations are relentlessly driving down the cost of a given unit of intelligence. This trend is set to continue, steadily lowering the barrier to entry and making powerful AI more accessible to smaller, innovative entrants.